AI augmented city-level solar panel scanning and geolocating from open data
Embedded Files
Shiliang Zhang, University of Oslo
Overview of this work
The dataset and part of the code can be downloaded at:
https://github.com/slzhang-git/GeoSolarDataAI
Our paper on this work: AI and Open-data Driven Scalable Solar Power Profiling
Our paper has been accepted by 18th IEEE PES APPEEC, 2026, Singapore
Pre-print version: doi: 10.48550/arXiv.2605.02738
Software, API, and demonstration
We release an API named Geosolar. Particularly, for any user‑specified building location, our API automatically retrieves the corresponding satellite imagery from open-source aerial imagery sources, detects rooftop solar panels, and returns georeferenced panel polygons that can be directly used in maps and programming works. We show how to use our API at our Github page. Below is an example of solar panel detection for a single building in Zurich, Switzerland.
We show an example of solar panel detection for the area Berg am Irchel in Switzerland in this html file, where we store all the geographic geometry of the detected panels at line 138 of the file. The visualization of the detected solar panels is shown below in an interactive map:
We show an example of solar panel detection for the city of Bülach in Switzerland, where we store all the geographic geometry of the detected panels. The visualization of the detected solar panels is shown below in an interactive map:
We show an example of solar panel detection for the whole island of Waiheke in New Zealand below in an interactive map (red: Solar panels on residential building rooftops. Yellow: solar panels on non-residential building rooftops):
Page updated
Google Sites
Report abuse